1. Field of the Invention
The present invention relates to a method for detecting the movement of an object by image processing, and a printer using the method.
2. Description of the Related Art
In printing while conveying media such as print sheets, if conveyance accuracy is low, density non-uniformity of a halftone image occurs, or a magnification error occurs. Therefore, the quality of an obtained print image is degraded.
Therefore, a printer includes a precise conveying mechanism using high-accuracy components. However, demands for higher printing quality are increased, so that further improvements in accuracy are desired. On the other hand, demands for cost are also increased. Thus, both the increase in accuracy and the reduction in cost are required.
In order to cope with this, attempts to image a surface of the medium and detect the movement of the conveyed medium by image processing have been made to detect the movement of the medium with high accuracy and realize stable conveyance by feedback control.
Japanese Patent Application Laid-Open No. 2007-217176 discusses a technique for imaging a surface of a moving medium more than once in time series at different timings by an image sensor, comparing obtained images by pattern matching processing, and detecting the movement amount of the medium.
In order to image the surface of the medium and detect the movement of the medium by image processing, it is required that a fine concavo-convex shape on the surface of the medium is sufficiently optically identifiable and a unique image pattern is clear.
When the surface of the medium to be used is significantly smooth, however, an image obtained by the imaging has a low contrast. Therefore, a unique pattern in the image may be difficult to identify. Thus, the accuracy of the pattern identification processing is degraded or the processing cannot be performed. This phenomenon will be described below.
FIG. 8 illustrates an example of images obtained by imaging a partial region of a surface of a moving medium at different timings. An image 101 is first captured, and an image 102 is captured after the medium moves by a predetermined distance. In both images, a fine concavo-convex shape on the surface of the medium (unevenness of paper fibers, etc.) appears as unique patterns having a high contrast.
It is determined where the same pattern as a partial region 101a indicated by a broken line in the image 101 exists in the image 102 based on the degree of similarity between the images by known pattern matching processing. The result of the determination shows that the pattern exists in a region 102a indicated by a broken line in the image 102. The movement amount 103 of the medium can be determined by detecting how many pixels are there between the regions 101a and 102a in a conveyance direction.
FIG. 9 illustrates an example of images 201 and 202 in a case where a medium whose surface has a concavo-convex shape smoother than that in FIG. 8. In the images, there are a fewer unique patterns formed by the concavo-convex shape on the surface of the medium than those in FIG. 8.
In the example of the image 201, a region 201a used for pattern matching processing includes no unique pattern. When an image pattern similar to the region 201a is searched for in the image 202, it is determined that a plurality of regions 203 are similar to the region 201a. Therefore, the movement amount of the medium cannot be accurately detected.
FIG. 10 illustrates examples of images 301 and 302 in a case where a medium whose surface has a concavo-convex shape smoother than that in FIG. 8. In the images, unique patterns are seen but have a significantly lower contrast than those in FIG. 8. If the contrast of the image is significantly low, information relating to the unique patterns greatly changes by only a slight change in the quantity of illumination light, for example.
A region 301a and a region 302a after movement may be actually recognized as different patterns by a variation in the quantity of illumination light before and after the movement, although they are originally the same unique pattern. Then, a matching cannot be detected by pattern matching processing. Therefore, the movement amount of the medium cannot be accurately detected.